Autocoder

Micron Interview Questions and Answers for India Software and Firmware Roles

Last updated 2026-11-19

Micron makes memory and storage, so a software or firmware interview there often starts from how memory actually works: why DRAM needs refresh, why NAND flash cannot be overwritten in place, how error-correcting codes protect data, and how you manipulate bits in C. This page gives twelve practice questions with worked answers across those areas, plus HR questions.

They are written for the topics such roles are commonly reported to cover, not taken from any interview. For problems to code, see Micron coding round questions.

Typical rolesFirmware, embedded software, test and validation software, SSD and memory-related roles
Core topicsMemory technologies, C, bit manipulation, OS, computer architecture, DSA
Rounds (commonly reported)Online test, technical interviews, HR

Rounds as commonly reported

Prep write-ups describe an online screening with C and technical MCQs, then technical interviews that mix C, data structures and subject basics, then an HR conversation. Micron does not publish a fixed process and teams differ, so rely on your own invitation for the exact structure.

Because the products are storage and memory, even a software question can arrive in storage clothing: how would you track which blocks are free, or how would you spread writes evenly? Practise connecting a data structure to a device constraint.

Topics to revise

Hiring status

This site holds no sourced hiring statements for Micron, so it quotes no drive dates, number of openings or packages. Treat posts that give them without a source link with caution.

Check the company's own careers site for current openings and how to apply.

Micron Interview Questions and Answers for India Software and Firmware Roles: questions with worked answers

  1. 1

    Compare DRAM and SRAM.

    easy

    A DRAM cell is one transistor and one capacitor, so it is small and cheap per bit, but the charge leaks and each row must be refreshed every few tens of milliseconds, and reading is destructive so the row is restored after access. An SRAM cell is typically six transistors forming a latch, so it is larger and costlier per bit, needs no refresh and is much faster. Hence SRAM is used for CPU caches and DRAM for main memory.

  2. 2

    Explain row hits and row misses in a DRAM bank.

    medium

    A DRAM bank reads a whole row into a row buffer (ACTIVATE), then serves column reads and writes from it, and finally closes it (PRECHARGE). If the next access is to the open row, it is a row hit and only needs the column access. If it is to a different row, the controller must precharge, activate the new row and then access, which takes several times longer. Controllers reorder requests and spread data over banks to increase row hits and bank parallelism.

  3. 3

    Why can NAND flash not be overwritten in place? Describe pages and blocks.

    medium

    NAND is programmed a page at a time (a few kilobytes) but can only be erased a whole block at a time (many pages), and programming can only change bits from 1 to 0; erasing returns the block to all 1s. To update data, the firmware writes the new version to a fresh page and marks the old page invalid. Erase cycles wear the cells out, so a block survives a limited number of program and erase cycles.

  4. 4

    What does a flash translation layer do? Explain wear levelling and garbage collection.

    hard

    The flash translation layer maps logical block addresses seen by the host to physical pages, which change on every write. Wear levelling spreads erases evenly by choosing less-worn blocks for new writes and occasionally moving cold data out of barely worn blocks. Garbage collection picks blocks with many invalid pages, copies the valid pages elsewhere and erases the block to reclaim space. The extra writes this causes are the write amplification, the ratio of data written to flash over data written by the host.

  5. 5

    How many parity bits does a Hamming code need for 8 data bits, and how does SECDED extend it?

    medium

    For m data bits you need r parity bits with 2^r >= m + r + 1. For m = 8, r = 4 works (16 >= 13), giving a 12-bit codeword. For 64 data bits r = 7 (128 >= 72). The syndrome computed from the parity checks gives the position of a single flipped bit, which can be corrected. Adding one overall parity bit makes SECDED: single-error correction, double-error detection, such as the (72, 64) code used in ECC memory.

  6. 6

    How would you track free blocks on a device with a million blocks using minimal memory?

    medium

    Use a bitmap with one bit per block: 1,000,000 bits is about 122 KB. To allocate, scan for the first zero bit, skipping 64 bits at a time by comparing a word with all ones, then set it. Finding a free block costs O(n / 64) at worst, and can be improved by remembering a hint index or by using a two-level bitmap. Freeing is clearing the bit in O(1).

    c
    #include <stdio.h>
    #include <stdint.h>
    #define WORDS 4
    static uint64_t map[WORDS];
    int alloc(void) {
        for (int w = 0; w < WORDS; w++) {
            if (map[w] != ~0ULL) {
                int b = __builtin_ctzll(~map[w]);   /* first zero bit */
                map[w] |= 1ULL << b;
                return w * 64 + b;
            }
        }
        return -1;
    }
    int main(void) {
        for (int i = 0; i < 70; i++) alloc();
        printf("%d\n", alloc()); /* 70 */
        return 0;
    }
  7. 7

    Explain wear levelling with a data-structure choice: how would you pick the block to write next?

    medium

    Keep an erase count per free block and always allocate the free block with the smallest count; a min-heap keyed on the erase count gives O(log n) allocation and O(log n) insert. To handle cold data that never moves, periodically swap a rarely changed block with a heavily worn one. The same idea, a priority queue ordered by a wear metric, appears in many schedulers.

  8. 8

    What is endianness, and why does it matter for a storage or network format?

    medium

    Little-endian stores the least significant byte at the lowest address, big-endian the most significant. Two systems must agree on the byte order of multi-byte fields in files, on-disk metadata and network packets (network order is big-endian). Write portable code by assembling values from bytes with shifts, as in (p[0] << 8) | p[1], instead of casting a byte pointer to a wider pointer, which also risks misaligned access.

  9. 9

    What is memory alignment, and what can go wrong if you ignore it?

    medium

    A type of size N is usually expected at an address that is a multiple of N. Misaligned access can be slower, can split across cache lines, and on some processors causes a fault. Casting a char buffer to a uint32_t pointer is a classic source of both alignment and aliasing problems; use memcpy into a local variable instead, which compilers turn into efficient code. Structures are padded to keep their members aligned.

  10. 10

    What is DMA and why do storage controllers use it?

    medium

    Direct memory access lets a controller move blocks between a device and memory without the CPU handling each word. The CPU sets the source, destination and length, and an interrupt signals completion. For storage that moves large pages continuously, it frees the processor for command processing. Take care with cache coherency, which needs invalidation or cleaning of the buffer, and with buffers that must be physically contiguous.

  11. 11

    Implement a queue using two stacks and state the amortised cost.

    medium

    Use an input stack for enqueue and an output stack for dequeue. When the output stack is empty on a dequeue, pop everything from the input stack and push it onto the output stack, reversing the order. Each element moves at most twice, so each operation is O(1) amortised, though a single dequeue can cost O(n). Common mistake: transferring on every dequeue, which gives O(n) each time.

  12. 12

    HR: Why Micron, and what interests you about memory and storage?

    easy

    Give a specific hook: a course on computer architecture, a project with an SD card or an SPI flash chip, or curiosity about how an SSD hides flash limitations. Say what you built and learned and what you want to work on. Avoid claims about the company that you cannot support from a source. Also see how to crack the HR interview.

Frequently asked questions

Do Micron interviews test hardware knowledge for software roles?

Commonly reported for firmware and embedded roles: memory types, flash behaviour and error correction can come up beside C and data structures. Pure application roles may be lighter on this.

How should I prepare for NAND flash questions?

Know pages versus blocks, why erase-before-write exists, what a flash translation layer does, and the ideas of wear levelling, garbage collection and write amplification.

Are these real Micron questions?

No. They are practice questions in the style of topics such roles commonly cover. Micron does not publish its interview questions.

Where do I check Micron openings?

Use the company's own careers site. This site has no sourced hiring facts for Micron, so it quotes none.

Related guides

More in Company Interview Questions

See all in this section